I know it's possible to get IDE autocompletion from the *Table classes in Doctrine by doing things like this:
SomethingTable::getInstance()-><autocomplete>;
But the most important part is missing. I want autocomplete on the model classes themselves, not just the Table classes. It appears that Doctrine is not properly declaring the PHPdoc #return object types in the find and other standard model methods.
For example what I want to be able to do is this:
$something = SomethingTable::getInstance()->find($id);
$something-><autocomplete>
and have that pop up the methods and properties of the Something class.
I should mention too that I don't specifically care about using the SomethingTable::getInstance() syntax at all. ANY decent syntax that's standard Symfony is acceptable. Most of the time I'm fetching objects (or Doctrine_Collections) via custom queries like this:
$somethings = Doctrine_Query::create()
->from('Something s')
->leftJoin('s.SomethingElse s2')
->where(...);
By the way, in case it's not clear, I'm asking if there's any automatic solution to this with ANY of the various Doctrine find, fetch or query syntaxes. I'm NOT asking how to manually edit all the PHPdoc headers to cause the behavior I want.
I'm using NetBeans 6.9.1 and Symfony 1.4.12 with Doctrine, but not everyone working on the same code uses NetBeans.
The problem is that autogenerated *Table classes have the wrong phpdoc #return in the getInstance() method:
/**
* Returns an instance of this class.
*
* #return object MyModelTable
*/
public static function getInstance()
{
return Doctrine_Core::getTable('MyModel');
}
You just need to manually fix the #return line deleting the word "object":
* #return MyModelTable
And magically IDE autocompletion just works, giving you all the instance and static methods:
MyModelable::getInstance()->... //(you'll have autocompletion here)
I know, its a pain to have to manually fix this but at least it only have to be done once for each model *Table file.
In netbeans its quite easy:
$foo = ModelNameTable::getInstance()->find(1); /* #var $foo ModelName */
/* #var $foo ModelName */ tells netbeans to handle the variable $foo as a ModelName class.
just fix the generated model files by adding
/**
* #return ModelNameTable
*/
in the comment of the getInstance() method. This will provide autocomplete for the model file.
Regarding the find method, you can edit the comment of the class like this :
/**
* #method ModelName find()
*/
I think it might be possible for you to do this automatically by creating you own skeleton files.
Or not : Symfony Doctrine skeleton files
You could use sed to achieve this, or perhaps build your own task using the reflection api.
Related
How to configure PhpStorm for visualizing available methods and attributes of an object variable while typing?
I remembered having seeing that adding the #var annotations on the line above of the declared variable as
// #var \Drupal\Core\Entity\Plugin\DataType\EntityAdapter $entityAdapter
$entityAdapter = $entityReference->getTarget();
permits it but I do not know what does it do and how to use it.
An example of autocompletion:
annotating a local variable in php
I am having trouble with getting functional tests to run in Symfony3 with Doctrine.
I have the code organized in two bundles with which need to be accessed by one EntityManager with entities stored in two different MySQL databases.
To achieve this, all Entities have a "schema" annotation in their definition, like this:
/**
* #ORM/Table(name="tablename", schema="schema")
* #Entity( ... )
*/
Without this setting, it has been my experience, that the Doctrine schema:create tool is not able to correctly create the entities in the right databases.
However it appears, that the schema annotation is not considered to be environment dependent.
So when I want to run functional tests that need to load fixtures, the ORMPurger tries to purge schema.tablename, where it should use the table/schema "test_schema".
Is there any way to keep the schema annotation but make it dependent on the environment, so that when the environment is "test", a different schema is used?
EDIT:
It appears that using the "schema" annotation for entities is pretty terrible all around when you are using different Symfony environments. At least when used in conjunction with MySQl, at least I think that that is the reason, since MySQL doesn't actually support schemas.
Every Symfony or Doctrine command I tried to take the schema annotation literally, regardless of the --env setting.
I've done some more digging and found what I needed to do perfectly laid out here:
Programmatically modify table's schema name in Doctrine2?
So I added an EventListener, that adds the correct schema according to the EM used so I don't need hard-coded schema annotations anymore.
Heres the code for the Listener I've made:
<?php
namespace /* ... */
use Doctrine\ORM\Mapping\ClassMetadata;
use Doctrine\ORM\EntityManagerInterface;
use Doctrine\ORM\Event\LoadClassMetadataEventArgs;
class MappingListener
{
public function loadClassMetadata(LoadClassMetadataEventArgs $eventArgs)
{
/** #var EntityManagerInterface $entityManager */
$entityManager = $eventArgs->getEntityManager();
/** #var ClassMetadata $classMetadata */
$classMetadata = $eventArgs->getClassMetadata();
$database = $entityManager->getConnection()->getDatabase();
$classMetadata->table['schema'] = $database;
}
}
I've started using Flight microframework, but all methods are hidden under the hood (not declared in the Flight class).
How can I configure PHPStorm or should I write new set of rules?
Update: use framework instance doesn't work
I've tried to use framework instance, but has no success — I have internal methods in the suggestion list:
Update: autocomplete implemented in the Flight framework
First of all: I'd suggest to submit new issue on their Issue Tracker asking to provide some sort of helper file (like below).. or implement it in any other way (e.g. via PHPDoc' #method for Flight class -- no helper needed and no changes in the actual code -- just PHPDoc) so that IDE (e.g. PhpStorm or Netbeans) would not complain for non-existing methods and you will have some code completion help from IDE.
Magic is good .. but not when whole interface is based on such magic.
On the actual question, which you can resolve yourself.
You will have to spend some time (half an hour or even less) and create some fake Flight class and put it anywhere in your IDE -- it will be used for code completion only. Yes, IDE may warn you about duplicate classes.. but that inspection can be turned off.
The idea is to create a class and declare all required methods as they should have been done if it would be an ordinary class. To start with (will resolve issues for first code example on their readme):
<?php
class Flight
{
/**
* Routes a URL to a callback function.
*
* #param string $pattern URL pattern to match
* #param callback $callback Callback function
* #param boolean $pass_route Pass the matching route object to the callback
*/
public static function route($pattern, $callback, $pass_route = false) {}
/**
* Starts the framework.
*/
public static function start() {}
}
Here is how it looks now:
As you can see Flight is underwaved -- IDE says that there is more than one class with such name in this project. Just tell PhpStorm to not to report such cases:
For adding methods to the original class via #method PHPDoc tags:
/**
* Bla-bla -- class description
*
* #method static void route(string $pattern, callback $callback, bool $pass_route = false) Routes a URL to a callback function
* #method static void start() Starts the framework
*/
class Flight
{
...
}
I have an autobean with a property that is only needed for the UI. I believe that you can null out values and the AutoBeanCodex will not serialized that property, but that equates to an extra step which is needed at serialization.
I was hoping for some annotation similar to the Editor #Ignore annotation. For example:
public interface Foo {
...
#Ignore
String getUiOnlyProperty();
}
So, other than nulling out the value at serialization time, is there any other way to keep an autobean property from being serialized?
Autobeans are meant to be a Java skin on a JSON/XML/whatever format - they aren't really designed to hold other pieces of data. That said, several thoughts that either nearly answer your question with out-of-the-box tools, or might inspire some other ideas on how to solve your problem.
You should be able to build read-only properties by omitting the setter. This isn't quite what you are asking for, but still might be handy.
Along those lines, the JavaDoc for the #PropertyName annotation seems to allude to this possible feature:
/**
* An annotation that allows inferred property names to be overridden.
* <p>
* This annotation is asymmetric, applying it to a getter will not affect the
* setter. The asymmetry allows existing users of an interface to read old
* {#link AutoBeanCodex} messages, but write new ones.
*/
Reading old messages but writing new ones seems like it might be closer to what you are after, and still allowing you to work with the thing-that-looks-like-a-bean.
The real answer though seems to be the AutoBean.setTag and getTag methods:
/**
* A tag is an arbitrary piece of external metadata to be associated with the
* wrapped value.
*
* #param tagName the tag name
* #param value the wrapped value
* #see #getTag(String)
*/
void setTag(String tagName, Object value);
...
/**
* Retrieve a tag value that was previously provided to
* {#link #setTag(String, Object)}.
*
* #param tagName the tag name
* #return the tag value
* #see #setTag(String, Object)
*/
<Q> Q getTag(String tagName);
As can be seen from the implementation of these methods in AbstractAutoBean, these store their data in a totally separate object from what is sent over the wire. The downside is that you'll need to get the underlying AutoBean object (see com.google.web.bindery.autobean.shared.AutoBeanUtils.getAutoBean(U) for one way to do this) in order to invoke these methods.
child class/interface decoded as a parent interface does not explode on decoding allowing the goods to be passed together before the marshalling steps. my immediate test of the actual code below is performing as expected.
public interface ReplicateOptions {
/**
* Create target database if it does not exist. Only for server replications.
*/
Boolean getCreateTarget();
void setCreateTarget(Boolean create_target);
//baggage to pass along
interface ReplicateCall<T> extends ReplicateOptions {
/**
* If true starts subscribing to future changes in the source database and continue replicating them.
*/
AsyncCallback<T> getContinuous();
void setContinuous(AsyncCallback<T> continuous);
}
}
I'm working on Magento templates, but this issue would apply to any template loading system.
As these templates are loaded by the template engine there's no way for the IDE (in this case Aptana) to know what object type $this is.
Potentially it could more than one object as a single template could be loaded by multiple objects, but ignoring this, what would the correct phpdoc syntax be to specify a specific class for the $this object?
You can define it like this:
/* #var $this type */
where type is a class name
To be clear, using $this should only ever indicate an object of the current class, right?
PhpDocumentor doesn't currently (v1.4.3) recognize $this as a specific keyword that should equate to a datatype of the class itself.
Only datatypes known by PHP and classes already parsed by PhpDocumentor are the proper datatype values to use with the #return tag. There is a feature request in to have some option available in PhpDocumtentor to aid in documenting fluent methods that always "return $this". [1]
In the case of the #var tag, I don't see how it would be feasible for a class variable to contain its own class instance. As such, I can't follow what "#var $this" should be saying.
If, however, your intention with $this is not for fluent methods that "return $this", and was simply to be some shortcut to PhpDocumentor and/or your IDE to magically guess what datatypes you might mean by using $this, I'd have to guess there's no way to do it. The closest suggestion I could make would be to use the name of a parent class that is a common parent to all the various child classes that this particular var/return might be at runtime, and then use the description part of the tag to have inline {#link} tags that list out the possible child classes that are possible.
Example: I have a Parent abstract class with Child1, Child2, and Child3 children that each could occur in my runtime Foo class.
So, Foo::_var could be any of those child class types at runtime, but how would I document this?
/**
* #var Parent this could be any child of {#link Parent}, {#link Child1}, {#link Child2}, or {#link Child3}...
*/
protected $_var;
Getting back to the "return $this" issue, I'd document things in a similar way:
/**
* a fluent method (i.e. it returns this class's instance object)
* #return Parent this could be any child of {#link Parent}, {#link Child1}, {#link Child2}, or {#link Child3}...
*/
public function foo() {
return $this;
}
Documenting this way at least allows your class doc to have links to the particular classes. What it fails to do is highlight the fluent 'ness. However, if your IDE is capable of recognizing the class names, then perhaps it will be able to do the necessary logical linking to those other classes. I think Eclipse is able to do this at least with popup help, if you hover over the class name in the tag's description. I do not think Eclipse can use this to then make the various child classes' methods available in code completion. It would know about the Parent methods for code completion, because the datatype I explicitly list is Parent, but that's as far as the IDE can go.
[1] -- http://pear.php.net/bugs/bug.php?id=16223
I have found that defining a type with #var for $this does not work - presumably because $this is special and is treated as such by Aptana. I have a similar need to the poster I think - it is in template files (in my case simply located and included by functions within the data class) that I wish to set a type for $this. As #ashnazg says, setting a type for $this within a class definition is not needed, because the type of $this is always the type of the class (up to inheritance).
There is, however, a workaround for template files. At the top of the template file simply put something like
/**
* #var My_Data_Model_Type
*/
$dataModel = &$this;
Then simply use $dataModel (or whatever you choose to call it - maybe something shorter) instead of $this in the template